Research On Key Technologies Of Condition-Based Maintenance For Mechanical Transmission Components Based On Markov Process

Equipment manufacturing level to determines the economic strength,national defense strength,and the ability to compete and cooperate of a country in the global economy.Accelerating the optimization and upgrading of the equipment manufacturing industry should be based on not only self-developed advanced manufacturing equipment,but also strict requirements for stable and reliable equipment operation.With the increase in automation of production processes,traditional breakdown maintenance and scheduled maintenance can no longer meet the actual needs of manufacturing enterprises.Benefiting from the development of information sensing technology,condition-based maintenance(CBM)can obtain state information of core components of manufacturing equipment in real time or quasi-real time by the condition monitoring system and make reasonable maintenance decisions based on the information.It lays an important foundation for ensuring the safety,reliability,and economy of equipment operation.As a key device to achieve power transfer and motion conversion in manufacturing equipment,the service performance of mechanical transmission components directly affects the reliability level of the system.The impacts of its failure include not only economic losses caused by production downtime,but also product quality fluctuations caused by performance degradation.Therefore,in actual production,there is an urgent need to assess the degradation level of mechanical transmission components accurately and promptly and adopt targeted proactive maintenance strategies to control operating costs while minimizing downtime.In this paper,supported by the national science and technology major projects,five key technical problems in the CBM of mechanical transmission components are studied along with the actual needs of manufacturing enterprises,including FMEA fault knowledge representation,signal preprocessing and feature reduction,health state assessment,information fusion,and maintenance strategy support.Guided by the fault knowledge accumulated in the reliability field test,the Markov process is introduced to establish the functional mapping relationship between monitoring information and performance degradation.Then,the established relationship is integrated into the maintenance strategy for mechanical transmission components to provide strong technical support to ensure the safe,smooth,costreducing and efficient operation of manufacturing equipment.This paper focuses on the following work:(1)To address the problem of how to utilize field fault data to obtain fault knowledge that can help guide the CBM process,an improved FMEA-based fault knowledge representation method for mechanical transmission components is proposed.Based on a comprehensive summary and analysis of the limitations of classical FMEA,a linguistic variable term set based on interval triangular fuzzy numbers is constructed for the representation and processing of uncertain information.For the weighting of risk indicators,a comprehensive weight calculation method based on ICWGT is proposed considering both subjective and objective weights;the risk ranking method for typical failure modes based on fuzzy VIKOR is extended.The application effectiveness of the proposed method is validated by a case of risk analysis of the servo turret power head transmission system.(2)To address the problem of how to construct a sensitive feature space that can characterize the degradation of mechanical transmission components in a strong noise background,a vibration signal feature space construction method based on adaptive VMD and manifold learning is proposed.Based on the comparison with WPT,EMD,EEMD and LMD,adaptive VMD is used to preprocess vibration signals.Time/frequency domain features and scale-domain features are extracted to form a high-dimensional feature space.A reduction process of high-dimensional feature space based on SL-Isomap is investigated to address the characteristics of high-dimensional feature space containing nonlinearity and redundancy.The proposed method is validated with the DDS planetary gearbox dataset,and an application analysis is performed in the state assessment of the servo turret power head transmission system.(3)Aiming at the problem that the health state assessment model cannot be established in the data accumulation stage of the whole life cycle,a health state assessment method based on GMM-HMM is proposed under the condition of incomplete data.Referring to Markov process,it is clear that the state transfer process of mechanical transmission components has Markov property.For cases with incomplete training data during the data accumulation stage,estimation of the component state is conducted by calculating the deviation between the testing sequence and the normal sequence.Referring to the plotted performance degradation curves,a health state discretization method based on the maximum inter-class-intra-class-ratio is proposed.Based on that,the health state assessment is transformed into a multi-class state identification problem using whole-life cycle training data with labels.The proposed method is validated with the XJTU-SY bearing dataset,and an application analysis is performed using the performance degradation test on the servo turret power head transmission system.(4)To address the problem of inaccurate assessment results due to uncertainty,incompleteness and imprecision of single-channel information,a multi-channel information decision-layer fusion method based on improved D-S evidence theory is proposed.The generation method of basic probability assignment is proposed according to the log-likelihood probability value of each channel output.For the problem of potential evidence conflict,a conflict metric considering the consistency of evidence is investigated as a way to calculate the weight between evidence.To obtain global fusion results,the basic probability assignments are weighted and averaged,and the weighted average evidence is combined using Dempster’s fusion rule.According to the probability assignment after fusion,the state transition matrix is calculated by introducing evidence Markov chain,which provides an effective basis for the formulation of maintenance strategy.The proposed method is validated by the case analysis in the previous section.(5)To address the problem that the existing ‘quality + maintenance’ conditionbased maintenance strategy models fail to take into account the operation plan,an Markov decision process condition-based maintenance strategy model for mechanical transmission components under the operation plan is developed.According to the operation plan of the manufacturing enterprise,the operation time is divided into planned production time and planned downtime.By analyzing the basic elements of the Markov decision process,a condition-based maintenance strategy model that integrates operations,quality,and maintenance is developed.Based on the established model,the value functions under different health states are derived and solved using the value iteration algorithm.In addition,the optimal maintenance strategy for planned production time with imperfect maintenance actions performed during planned downtime is further discussed.The proposed method is applied and analyzed through the simulation case of maintenance strategy modelling for the servo turret power head transmission system.Through the study of the above key technologies,this paper finally constructs a framework for the condition-based maintenance of mechanical transmission components,which solves the problem of the lack of an effective interface between theoretical research and engineering practice.It provides practical and reliable technical guidance for manufacturing enterprises to implement condition-based maintenance.